JP2582611B2 - How to create a multi-font dictionary - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] Regarding a method for creating a multi-font dictionary for recognizing a plurality of character fonts with a single dictionary, it recognizes many types of fonts with a high recognition rate even with a single dictionary It is an object of the present invention to provide a dictionary creation method that can read a character to be recognized with M (M is an integer) binarization thresholds and obtain a contour of a character composed of the binarized pixels. , A step of dividing a one-character area into N × N (N is an integer) sections, and a step in which pixels constituting a contour line in each section are sequentially traced with respect to a next adjacent pixel as a reference direction. From
A process of creating an N × N × 4 dimensional feature amount for each stage of each character by indexing corresponding to four directions at intervals of 45 °, and a plurality of features obtained for each of a plurality of character fonts. And generating a feature amount obtained by averaging the feature amounts of the character fonts. A feature amount group obtained by character samples including a predetermined number of different characters is used as a common dictionary for the plurality of character fonts.

[Industrial applications]

The present invention relates to a method for creating a multi-font dictionary for recognizing a plurality of character fonts with a single dictionary.

[Conventional technology]

In the type recognition apparatus, a document is read via a scanner, and printed characters are identified and coded one by one with reference to a dictionary to create a document file. Further, a system for performing various operations such as displaying a document or processing a document in combination with a word processing function can be configured.

The dictionary used in such a type recognition device is:
It is desired that even if there are many types of character fonts of the type to be recognized, the type can be identified at a high recognition rate.

When performing type recognition (here, “type” refers to a kanji code defined by JIS), standards are defined by JIS for target character fonts.
However, in reality, print fonts in newspapers, magazines, etc., and print fonts in word processors, printers, etc., are similar to the fonts specified in JIS, but differ in detail depending on the manufacturer. .

Therefore, the conventional type recognition device has a dedicated dictionary corresponding to each of these various fonts,
Thus, type recognition was performed.

In a type recognition device, it is desirable to be able to handle documents composed of various different fonts in common. In this case, however, preparing a dictionary for each font requires a huge amount of data and is practical. Not.

Therefore, it is conceivable to use a single dictionary to recognize many types of fonts.However, in this case, there is a difference in fonts between the fonts as described above. If the user tries to recognize different character fonts using the dictionary created in this way, the recognition rate may deteriorate. Therefore, conventionally, there was a problem that it was difficult to recognize many types of fonts using a single dictionary.

[Problems to be solved by the invention]

SUMMARY OF THE INVENTION The present invention is to solve such a problem of the prior art, and provides a dictionary creation method that can recognize a variety of fonts at a high recognition rate even with a single dictionary. It is intended to be.

[Means for solving the problem]

As shown in FIG. 1, the method of creating a multi-font dictionary according to the present invention reads pixels obtained by binarizing a character to be recognized with M gradations (M is an integer). A step of extracting a contour of a character composed of binarized pixels (steps S2, S3, S4); and a step of dividing an area of one character into N × N (N is an integer) sections (step S5). By indexing the direction with respect to the next adjacent pixel when sequentially tracing the pixels constituting the contour line in the segment in correspondence with four directions at 45 ° intervals from the reference direction, N × N × 1 for each gradation of each character
Step of creating N × 4 dimensional features (steps S6 and S7)
And a step of creating a feature amount by averaging a plurality of feature amounts obtained for each of a plurality of character fonts (step S8)
And a feature dictionary obtained by a character sample composed of a predetermined number of different characters is used as a common dictionary for the plurality of character fonts.

Accordingly, the configuration of the present invention is as described below. That is, a character to be recognized is read with M (M is an integer) binarization thresholds, and a contour of a character composed of the binarized pixels is extracted. Is an integer) by dividing the image into sections, and by indexing the directions to the next adjacent pixel when sequentially tracing the pixels constituting the contour line in each section, corresponding to four directions at 45 ° intervals from the reference direction. A step of creating an N × N × 4 dimensional feature for each stage of each character; and a step of creating a feature by averaging a plurality of features obtained for each of a plurality of character fonts. It has a feature as a method for creating a multi-font dictionary, wherein a feature amount group obtained by a character sample including a predetermined number of different characters is used as a common dictionary for the plurality of character fonts.

[Action]

In the multi-font dictionary creation method of the present invention, the first
As shown in the figure, first, a sentence to be a dictionary is read as one screen from a reading device such as a scanner (step
S1) Then, this is converted into a screen composed of a set of binary pixels of black and white corresponding to the shading of the image using an appropriate threshold value (step S2). Next, such a sentence image is divided and cut out for each character line, and each line is further divided and cut out for each character constituting the line (step S3).

A character outline is extracted from each character image thus separated (step S4). Extraction of a character outline can be performed by a known method such as detection of a boundary of white-to-black or black-to-white change of a pixel when a screen is scanned in a certain direction.

Next, the area of one character is equally divided into N × N squares (step S5). When the pixels on the contour line in each cell are sequentially traced from the top along the contour line, the direction with respect to the next adjacent pixel is four directions of 45 ° every 45 degrees corresponding to the angle from the reference direction. Indexing is performed in four directions, each of which is indicated by an index indicating the index (Step S6).

FIG. 2 shows indices in four directions in step S6 of the method for creating a multi-font dictionary according to the present invention, where the right horizontal direction is set to 0, and then 1,2,3 every 45 degrees counterclockwise.
It is assumed that an index in four directions is defined. Therefore, for example, when the direction to the next adjacent pixel is vertically downward, the exponent is 2.

When indexing is completed for all pixels on the contour line in a certain grid in this way, the number is added for each of the same index components, and the numbers are arranged in ascending order from the reference direction to form a four-dimensional feature. Amount. For example, if the feature value is (2,0,
When expressed as (0,3), the number of exponent "0" is 2, the number of exponent "1" is 0, the number of exponent "2" is 0, and the number of exponent "3" is 3 Is shown. Therefore, the feature amount when no pixel exists in the cell is indicated by (0,0,0,0).

By performing such an operation for each square constituting the character image and creating a four-dimensional feature amount for each square, a feature amount of N × N × 4 dimensions as a whole is created for each character. (Step S7).

Such processing is performed M times while changing reading conditions such as density. That is, starting from the reading number i = 0, i>
Iteratively repeats while incrementing i by 1 until it reaches M,
Thus, a feature amount group a is created.

Further, in accordance with the method of creating the feature amount a, the same sentence of the K fonts is read and the same processing is performed. That is, starting from the font number j = 1 until j> K, j
Are sequentially repeated while adding +1 to obtain feature amount group b, feature amount group c,.

When the creation of the K types of feature amount groups a, b,... Has been completed, a desired dictionary is obtained by averaging the feature amounts of all groups. That is, data obtained by averaging the feature amounts of each character is obtained by adding the feature amounts of each group and dividing the sum by K × M. This is used as dictionary data of each character.

FIG. 3 shows an example of the configuration of the multi-font dictionary creating unit 10 corresponding to the principle configuration of the method for creating the multi-font dictionary of the present invention shown in FIG. That is, the document input is binarized and temporarily stored in the image memory 11.
Next, the character outline extraction unit 12 performs processing such as line extraction, character extraction, and character outline extraction based on the data stored in the image memory 11. The feature amount generation unit 13 divides N × N squares into squares based on the data stored in the image memory 11 and the character outline data created by the character outline extraction unit 12 to calculate each pixel constituting the character outline in the square. A plurality of feature amount groups corresponding to each character font are created by performing processing such as N × N × 4 dimensional feature amount creation by indexing in four directions with respect to the direction of the next adjacent pixel, and furthermore, the feature of each group The dictionary data is created by averaging the quantities. The created dictionary is output and stored in the external storage device.

Embodiment FIG. 4 shows a processing flow of a method for creating a multi-font dictionary as one embodiment of the present invention.

FIG. 5 shows an example of a target character font in this case.

In the embodiment shown in FIG. 4, a document (character sample) in which 3303 JIS kanji codes are arranged in code order as a document to be read from the scanner is used, and M = 10 is shown. The kind of character font used in this case is K = 3, that is, as shown in FIG. 5, font A (JIS old font) used for FM16β printer (Fujitsu), FMR printer (Fujitsu) Fonts B (JIS new fonts) and Iwata Hoshimitsu (standard typesetting fonts) are used for fonts C. Therefore, a total of K × M = 30 times are processed and averaged. Create a dictionary.

The processing is performed according to the flow shown in FIG. In this case, each step is indicated by the same number corresponding to each processing step shown in FIG. That is, binarization of read characters, line segmentation, character segmentation, character outline extraction, and the like are performed to divide into 8 × 8 squares to create 8 × 8 × 4 = 256-dimensional feature amounts. Further, such a process is repeated 10 times while changing the reading density or the like to obtain a feature amount group, and a similar process is performed on three types of fonts, thereby performing a total of 30 processes.

When all the data is obtained, dictionary data of each character is created by adding each feature amount group, dividing by 30 and averaging. Further, by performing such processing for each character constituting the character sample, a dictionary is created in which the character code is associated with the dictionary data of each character.

This dictionary associates each character constituting a character sample with an averaged feature amount of each character. When performing type recognition, a fourth character is read from the read character.
Character outlines extracted by the same processing as shown in the figure, the characteristic amount generation and the like, and the characteristic amount of each character obtained are collated with the above-described dictionary to select the same or closest characteristic amount. Type recognition can be performed by reading a character code corresponding to the feature amount from the dictionary.

When the above-mentioned character samples were evaluated using this dictionary, a recognition rate of around 99% was obtained. This recognition rate sometimes exceeded the recognition rate when a dictionary was created for each font and evaluated.

FIG. 6 shows a configuration example of a type recognition apparatus to which the multi-font dictionary creation method of the present invention is applied.
Is a multi-font dictionary creating unit shown in FIG. 3, and 14 is a scanner. 15 is a personal computer, a central processing unit (CP
U) 16 and a main storage device 17 for storing an operation program of the CPU 16, operation data, and the like. Reference numeral 18 denotes an external storage device.

In FIG. 6, the scanner 14 reads a document into the multi-font dictionary creator 10, whereby the multi-font dictionary creator 10 extracts character outlines, creates feature amounts, etc. as described with reference to FIG. Is performed. The personal computer 15 takes out the feature amount for each character constituting the character sample and extracts it from the external storage device
To memorize.

In this way, when the processing is performed M times for a certain character font, and the same processing is repeated for K character fonts of the same character sample, K × M times of processing is completed, the personal computer 15 becomes an external storage device. The dictionary data of each character is created by taking out the data of all the feature amount groups from 18 and adding and averaging them by K × M. The created dictionary data is output to the external storage device 18 and stored. By performing such processing for each character constituting the character sample, a dictionary for the character sample is created.

〔The invention's effect〕

As described above, according to the present invention, a dictionary for performing type recognition by averaging feature amounts extracted from a plurality of character fonts is created, so that a dictionary emphasizing the common part of each character font is created. Can be obtained, so that a single dictionary can be used for recognizing printed characters composed of various types of character fonts at a high recognition rate.

[Brief description of the drawings]

FIG. 1 is a diagram showing a basic configuration of a method for creating a multi-font dictionary of the present invention. FIG. 2 is a diagram showing 4-directional indices in step S6 of the method for creating a multi-font dictionary of the present invention. FIG. 4 is a diagram showing a configuration example of a multi-font dictionary creation unit corresponding to the principle configuration of the multi-font dictionary creation method of the present invention shown in FIG. 4; Fig. 5 is a diagram showing a processing flow of the method, Fig. 5 is a diagram showing an example of a character font in a method for creating a multi-font dictionary as an embodiment of the present invention, and Fig. 6 is an application of the method for creating a multi-font dictionary of the present invention. It is a figure showing the example of composition of a type recognition device. 10 Multi-font dictionary creation unit 11 Image memory 12 Character contour extraction unit 13 Feature generation unit 14 Scanner 15 Personal computer 16 CPU 17 Main storage device 18 External storage Equipment S1 ~ S8 …… Step

Claims (1)

(57) [Claims] A step of reading a character to be recognized with M (M is an integer) binarization thresholds and extracting a contour of a character composed of the binarized pixels; Dividing into N (N is an integer) sections; indexing the direction to the next adjacent pixel when sequentially tracing the pixels forming the contour line in each section in correspondence with four directions at 45 ° intervals from the reference direction A step of creating an N × N × 4 dimensional feature for each stage of each character, and creating a feature by averaging a plurality of features obtained for each of a plurality of character fonts. A method for creating a multi-font dictionary, wherein a feature amount group obtained by a character sample composed of a predetermined number of different characters is used as a common dictionary for the plurality of character fonts.